Liquid droplet germ granules require assembly and localized regulators for mRNA repression

Cytoplasmic RNA-protein (RNP) granules have diverse biophysical properties, from liquid to solid, and play enigmatic roles in RNA metabolism. Nematode P-granules are paradigmatic liquid droplet granules and central to germ cell development. Here we analyze a key P-granule scaffolding protein, called PGL, to investigate the functional relationship between P-granule assembly and function. Using a protein-RNA tethering assay, we find that reporter mRNA expression is repressed when recruited to PGL granules. We determine the crystal structure of the PGL N-terminal region to 1.5 Å, discover its dimerization and identify key residues at the dimer interface. In vivo mutations of those interface residues prevent P-granule assembly, de-repress PGL-tethered mRNA and reduce fertility. Therefore, PGL dimerization lies at the heart of both P-granule assembly and function. Finally, we identify the P-granule-associated Argonaute WAGO-1 as crucial for repression of PGL-tethered mRNA. We conclude that P-granule function requires both assembly and localized regulators.

λN22 provides tethering. PGL::SNAP::λN22 homozyotes were sterile (100%, n=94), but could 105 be maintained and tested as a fertile heterozygote (PGL-1::SNAP::λN22/+). Given that pgl-1 null 106 homozygotes are fertile 21 , the fertility defects from the addition of λN22 to PGL-1 is likely not 107 due to defective protein function. Regardless, the logic of our tethering strategy is simple: if 108 tethered PGL-1 localizes to granules and represses GFP expression as predicted, this assay

145
To test the significance of P-granule assembly to mRNA repression, we sought to perturb PGL-146 1 assembly into granules. An emerging principle is that multivalent-multivalent interactions drive 147 granule formation (e.g. protein with at least two multimerization domains) 1,33 . Because PGLs 148 are key assembly proteins for P-granules and can form granules on their own 6, 11, 24, 25 , we 149 reasoned that PGLs use multiple self-interactions to drive granule assembly. We previously 150 identified one dimerization domain (DD) centrally in PGL 27 , but DD missense mutations grossly 151 affected protein stability. We postulated the existence of another PGL multimerization domain 152 that might be more amenable to manipulation and again turned to structural analyses.

154
The region N-terminal to DD has high sequence conservation (Figure S1B), which implies a 155 critical role in PGL function. Our initial efforts to express trypsin-mapped recombinant protein fragments 27 of this N-terminal region proved unfruitful. However, addition of amino acids 157 disordered in the DD crystal structures 27 permitted robust expression sufficient for biochemical 158 and structural characterization (Figure S4A,B, see Methods for more details). Henceforth, we 159 refer to this stable protein fragment as the N-terminal dimerization domain (NtDD) (Figure 1B).

160
We determined the C. japonica PGL-1 NtDD crystal structure to 1.5 Å (Figure 2, see Table S1 161 for statistics, see Methods for details on crystallization and structure determination). The NtDD 162 had a novel fold consisting of 11 alpha helices and a single N-terminal beta strand (Figure 2B).

163
The asymmetric unit (ASU) was composed of four NtDD domains (Figure 2A

168
The complexity and conservation of these interactions suggested biological relevance. We next

173
These analyses yielded two distinct mutants: R123E with a single mutated residue and K126E 174 K129E with two mutated residues. Both NtDD interface mutants formed monomers rather than 175 dimers in solution (Figure 3D,E). We conclude that the dimers observed in the crystal structure 176 represent the NtDD dimer detected in solution. Therefore, PGL proteins possess two 177 dimerization domains. We renamed the original DD to Central DD (CDD) for clarity (Figure 1B To assess the role of NtDD dimerization in PGL granule self-assembly, we used an assay in 181 mammalian cells where expression of wild-type PGL-1 tagged with GST was sufficient for assembly into granules 25 . Similar to that report, wild-type PGL-1 tagged with GFP also formed 183 large cytoplasmic granules when expressed in mammalian cells (Figure 3F,G), while GFP 184 alone was diffuse ( Figure 3H). However, PGL-1::GFP mutated to either K126E K129E or 185 R123E no longer self-assembled into granules (Figure 3I,J). We conclude that NtDD 186 dimerization is essential for granule formation in mammalian cells.

188
To assess the role of NtDD dimerization in nematode germ cells, we used CRISPR to introduce  (Figure S5G,H). Therefore, both PGL-1 mutant 215 proteins are capable of incorporating into granules, but do so much more inefficiently than their 216 wild-type counterparts ( Figure 5C).

255
PGL tethering provides a simple assay for identification of additional factors needed for P-256 granule mRNA repression ( Figure S7A). We depleted candidate P-granule-associated RNA 257 regulators with RNAi and sought GFP reporter de-repression in PGL-1::SNAP::λN22 worms. In 258 this candidate screen, knockdown of the cytoplasmic Argonaute WAGO-1 had a dramatic effect 259 ( Figure S7B). RNAi against other candidates, by contrast, had either no or a minor effect on GFP repression ( Figure S7B). The RNAi screen highlighted WAGO-1 as a key factor in PGL-261 mediated mRNA repression, a finding consistent with previous studies showing that WAGO-1 262 localizes to P-granules and regulates gene expression 37, 38 .

264
To further investigate wago-1, we first inserted an epitope-tag at the endogenous locus ( Figure   265 S7C, Methods). WAGO-1::3xV5 co-localized with PGL-1(wt)::SNAP in perinuclear granules 266 ( Figure 6A), consistent with a previous report that WAGO-1 resides in P-granules 37 . We next 267 asked if WAGO-1 association with perinuclear granules was dependent on PGL assembly. In   (Figures S9E-I and S10). We conclude that WAGO-1 is a 293 major regulator of mRNA repression in P-granules. These results also provide direct evidence 294 that granule formation alone is not sufficient for mRNA regulation (see Figure 8, Discussion).

298
This work investigates the functional relationship between assembly of a paradigmatic liquid 299 droplet, the C. elegans P-granule, and the activities of its regulatory components. Our analyses 300 make three key advances. First, we discover that PGL-1 dimerization of the N-terminal domain 301 is required for its assembly into P-granules. Second, we demonstrate that PGL-mediated mRNA 302 repression relies on PGL granule assembly. Third, we find that mRNA repression by P-granules 303 employs the activity of at least one P-granule constituent, the Argonaute WAGO-1. Together,

304
these advances support a model that PGL assembly into granules is necessary for its biological 305 function but not sufficient to repress the expression of localized mRNAs (Figure 8). Below we 306 discuss these advances and their implications for RNP granules more generally.

308
Dimerization drives PGL protein assembly into granules 309 An emerging principle of RNP granule assembly is that multivalent macromolecular interactions 310 drive granule formation 1,33 . This work extends that principle to nematode P-granules and their 311 primary assembly proteins, the PGLs (see Introduction). A previous study reported that PGL proteins possess a dimerization domain (DD) in their central region 27 , and here we report the 313 discovery of a second PGL DD in the N-terminal region (NtDD). Thus, PGLs possess two 314 protein folds that confer multivalency. Based on insights from the NtDD structure, we designed 315 two distinct PGL-1 mutant proteins that are unable to dimerize in vitro and are severely 316 compromised in vivo for assembly into P-granules and fertility. These results provide evidence 317 that PGL dimerization is crucial for P-granule assembly and for P-granule biological function.   acids required for PGL assembly into granules (see above) allowed us to compare RNA 354 regulation when coupled to assembly-competent or assembly-defective PGL. We found that a 355 reporter mRNA tethered to assembly-competent PGL protein localized with PGL to P-granules 356 and was repressed, but a reporter tethered to an assembly-defective PGL did not localize to 357 granules and was expressed. Together, these results suggest that localizing mRNAs into P-   Figure 8C). Therefore, the scaffolding 387 function of PGL is important but not sufficient for P-granule function. We also propose that,

400
The functional relationship between PGL and WAGO-1 is unknown. PGL may simply recruit or 401 retain mRNAs in P-granules to be regulated by WAGO-1. A more enticing model is that PGL

456
japonica PGL-1 also crystallized as large (60-150 Å) rhomboid crystals in 40-45% PEG 400 at low (Na 457 Citrate pH 5.5-6.0) and physiologic pH (imidazole pH 7.5-8.0). Crystals grown in citrate or imidazole both 458 diffracted well, but we used imidazole (100 mM imidazole pH 7.5, 45% PEG 400, 1 mM TCEP pH 7.4) 459 due to its higher reproducibility for large crystals and its modestly better resolution. The crystals did not 460 require additional cryo-protection due to the high PEG 400. We eventually collected a full data set to 1.5 461 Å in space group C2.

539
with the co-injection marker phenotype were additionally screened by a combination of PCR without or 540 with restriction enzyme digest to identify those with the repair of interest. In JK5687, a SNAP tag 30 was 541 inserted between PGL-1 amino acids G713 and G714 in N2 worms. A 3xMYC tag was added to the N-542 terminus of GLH-1 between G17 and F18. A 3xV5 tag was added in the C-terminal region of PGL-3 543 between residues G627 and S628. A 3xV5 tag was added to the C-terminus of WAGO-1 between 544 residues E914 and A915. To generate the wago-1 null allele, a WAGO-1::3xV5 allele was mutated so that 545 648 base pairs were deleted from the N-terminus and proper coding frame shifted to add premature stop 546 codons ( Figure S7C). The wago-1 null allele was confirmed by staining and imaging (Figure 6D). F2s 547 were PCR screened to identify homozygous SNAP alleles and the PCR product sequenced to confirm 548 proper repair. Three worm strains were too infertile to freeze. All worms were outcrossed at least twice

552
Worms were singled into the peripheral wells of a 24-well plate that contained NGM agar and OP50 553 bacteria. Worms were allowed to propagate for 5 days at 20°C or 25°C, and then scored for progeny and 554 gravid progeny. We report the progeny numbers here.

626
We observed that the Surface-Surface colocalization created a new set of colocalization surfaces, which 627 are differently shaped and larger than the surfaces written with Imaris algorithms to detect gfp or 628 PGL:SNAP surfaces. Therefore, the total gfp intensity in the coloc surfaces was sometimes larger than in 629 the gfp surfaces.

640
Each of the two independent samples were run through a two-tailed t test to calculate significance. We                   fixed, stained and imaged in same region of meiotic pachytene (see Figure S5A). (C-F)

878
Representative images of SNAP staining to visualize PGL-1 expression and granule formation.

879
All images are partial z-stacks to maximize visualization of P-granules. Images were taken from          Number of TLS groups 1 1 Statistics for the highest-resolution shell are shown in parentheses.